|Publication number||US5390315 A|
|Application number||US 07/898,874|
|Publication date||Feb 14, 1995|
|Filing date||Jun 15, 1992|
|Priority date||Jun 15, 1992|
|Also published as||EP0574692A1|
|Publication number||07898874, 898874, US 5390315 A, US 5390315A, US-A-5390315, US5390315 A, US5390315A|
|Inventors||Geoffrey O. Blandy, Catherine M. Rossi|
|Original Assignee||International Business Machines Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (10), Non-Patent Citations (6), Referenced by (40), Classifications (12), Legal Events (5)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The invention relates generally to computer systems, and deals more particularly with a technique for allocating locations on DASD to store data; the allocation efficiently uses available storage in the DASD and minimizes overhead associated with accessing of the data on DASD.
In many computer systems, such as an IBM VM/ESA 1.1 computer system, external storage is provided by a direct access storage device (DASD). The DASD can comprise one or more magnetic disks. If more than one magnetic disks is included in the DASD, they are stacked as "platters" along a common axis. Each of the disks comprises a large number of coaxial circular tracks to store the data. If the DASD comprises a single disk, then a single read/write head is provided. The disk rotates such that when the read/write head is stationary, the read/write head can read all the information stored on a track during one complete rotation of the disk. The read/write head can also move radially to read other tracks. If the DASD comprises multiple disks, then one read/write head is provided for each disk. All the read/write heads are aligned under the same location of each disk and fixedly connected to each other to move in unison. Only one read/write head is sensed at any one time. A "cylinder" is defined as a collection of circular tracks, one per disk, that are vertically aligned. For a DASD comprising multiple disks, an extent comprises one or more cylinders. To read one cylinder, the DASD is rotated a number of times equal to the number of platters, and during each rotation, a different one of the read/write heads is sensed. To read one extent of the type comprising multiple adjacent cylinders, the foregoing process is repeated for each cylinder with a corresponding radial movement of the read/write head.
In the foregoing DASD arrangements, one factor which limits the speed of accessing (writing to or reading from) the DASD, is the time required to move the read/write head(s) radially from cylinder to another. Therefore, it is desirable to minimize the amount of radial movement of the read/write head required for each access. If the allocated addresses are not contiguous, then the allocated addresses may be radially spaced from each other, and require additional radial movement of the read/write head between the spaced addresses. A second factor which limits the speed of accessing the DASD is the time required for the data to rotate so that it is positioned under the read/write heads. Therefore, it is desirable to minimize the amount of rotational movement required for each access. When I/O is performed to non-contiguous addresses, extra time is required to allow the data that is not being accessed to pass under the read/write heads. A third factor is the number of I/O operations required to complete the access. An I/O subsystem controls the DASD and must specify a start location on DASD and length information for each discontiguous set of storage locations. Therefore, it is desirable to minimize the number of I/O operations that are required for each I/O access. If the data to be written to DASD can all be stored in one contiguous set of locations in DASD, then a single I/O operation suffices because only one start location and one length need be specified. However, if there is not one contiguous set of locations which is available to accomodate all the data, then the data must be split between discontiguous of locations, and this requires one I/O operation for each discontiguous set.
A fourth factor in managing the DASD is to avoid waste of storage capacity. Unfortunately, during continued usage of the DASD the available storage areas may become fragmented. For example, assume that ten pages of data (4,096 bytes per page) are stored in a fully utilized area of DASD. Each page utilizes one "slot" of storage. Next, an application program requests access to the ten pages, and the ten pages are written into main storage (RAM) to facilitate such access. The application program modifies eight of the ten pages and adds a page, but does not change the other two original pages. Subsequently, the operating system detects that the main storage is nearly full, and initiates a "block paging" operation to free-up the main storage by writing some of the pages in the main storage to DASD. There are various criteria by which the operating system decides which pages to write to DASD to free-up the main storage. According to one common criterion, the operating system selects a group of pages that were accessed at approximately the same time. This may lead to the selection of all of the foregoing ten pages. However, to minimize the amount of I/O work, the two unchanged pages are not written back to DASD; the original copy on the DASD is still valid. Therefore, only the eight updated pages and the additional page must be written to DASD, but the original location in the congested area can only store eight pages in contiguous locations. Therefore, a new location must be selected to store some or all of the nine pages. If a single set of nine contiguous slots is sought to minimize the number of I/O operations and limit radial movement of the read/write head, there will be an isolated set of eight contiguous slots which is available for other data. This process frequently repeats so that over time, there may be many isolated sets of different numbers of available slots on the DASD. To avoid wasting available storage on the DASD, techniques must be used to either avoid creating such isolated sets or to locate and use them for subsequent write operations. However, as noted above, the number of I/O operations and amount of radial head movement and rotational delay should be minimized.
One prior art technique logically divides the DASD into sets of slots of a fixed size, for example, ten slots per set. Whenever the operating system needs to write data to DASD, one or more of such sets is allocated to accomodate the request. Because of other criteria which govern the amount of data to be written to DASD, the amount of data to be written is often less than the allocation. This wastes storage capacity and leaves gaps which require rotational movement to bypass. Also, if the data to be written requires more than one set of the slots, then two or more I/O operations are normally required. The two or more sets are selected by a "moving cursor" which steps through the DASD in ascending DASD address to satisfy each successive request.
Another prior art technique utilizes a moving cursor to step through the DASD in ascending DASD address to satisfy each storage request made by the operating system. According to this technique, the allocations are not restricted to sets of fixed size, but instead attempt to precisely match the length of the storage request. As the moving cursor proceeds through the DASD to satisfy new requests, some of the data previously written to the DASD is accessed, i.e. read into main storage. Then a subset of those pages may be modified or regrouped and rewritten to a new DASD address, freeing the old DASD locations. This leaves gaps in the DASD. Another operating system function constantly monitors the availability of storage behind the cursor, and when seventy five percent of the slots behind the cursor become available, the cursor is moved back to the beginning. Then, the cursor moves forward again and allocates the available slots in order of location on the DASD to accomodate subsequent requests. While this technique minimizes the amount of available sets of isolated slots, the technique tends to fragment the storage of related data requiring extra I/O operations and radial and rotational movement of the read/write head.
Accordingly, a general object of the present invention is to provide a computer system, program or method which allocates DASD to minimize the number of I/O operations required to accomodate the data to be written to DASD, yet efficienlty utilizes the available storage on the DASD.
Another object of the present invention is to provide a computer system, program or method of the foregoing type which minimizes the amount of radial read/write head movement and rotational movement required to access the data.
The invention resides in a computer system, method and program product for allocating locations in DASD to store a specified length of data. The system, method and program product record addresses and lengths of sets of contiguous available storage locations in the DASD. Then, the system, method and program product searches the record to attempt to identify a set of contiguous available storage locations having the specified length and another set of contiguous available storage locations having a length which exceeds the specified length by an amount likely to accomodate most lengths of data subsequently to be stored on the DASD. If the search is successful in identifying a set of contiguous available storage locations having the specified length, the set is allocated to store the data, and if the search is not successful in identifying a set of contiguous available storage locations having the specified length but is successful in identifying another set of contiguous available storage locations having a length which exceeds the specified length by an amount likely to accomodate most subsequent lengths of data to be stored in the DASD, then the other set is allocated. To minimize radial movement of the read/write head, only sets that are located within the same extent may be recorded.
According to one feature of the present invention, the system, method and program product also attempt to identify a third set of contiguous available storage locations having a length which exceeds the specified length by a lesser amount than any other set in the record, and if the search for the set whose length equals the specified length and the search for the set whose length exceeds the specified length by an amount that is likely to accomodate most subsequent lengths of data are both unsuccessful, then the third set is allocated.
According to another feature of the present invention, the search also attempts to identify from the record two or more sets which together have a length equal or greater than the specified length if no single set in the record has a length which equals or exceeds the specified length. In such a case, the two or more sets are allocated.
FIG. 1 is a block diagram of a computer system which incorporates the present invention.
FIG. 2 is a pictoral representation of a slot allocation buffer within the computer system of FIG. 1, which buffer indicates the locations and sizes of available sets of contiguous slots within each extent.
FIGS. 3-9 form a flow chart of a slot allocation routine within the computer system of FIG. 1 which routine allocates DASD to accomodate the data to be written from main storage.
Referring now to the figures in detail wherein like reference numerals indicate like elements throughout the several views, FIG. 1 illustrates a computer system, generally designated 100, which includes the present invention. Computer system 100 comprises a CPU 110, main storage 115, DASD 160, an I/O subsystem 162 and an operating system 140. By way of example, CPU 110 and associated hardware (not shown) are provided by an IBM System/390 computer system. This system is described in a document entitled ESA/390 Principles of Operation, which document is available from International Business Machines Corp. at Mechanicsburg, Penna. by order number SA22-7201-00. Virtual storage 165 is allocated to each user/application program to store the application program and associated files or other data. The contents of virtual storage may be stored in resident virtual storage 167 in main storage 115, DASD 160 or elsewhere. Main storage 115 stores the operating system 140, associated tables and control blocks generated during execution of the operating system, and data and pages of virtual storage during manipulation by the CPU. DASD 160 provides a large amount of nonvolatile storage for the files and other data generated by the users/application programs and to hold the contents of virtual storage pages when not being accessed by the CPU. The I/O subsystem provides low level control for accessing specified locations in the DASD. The operating system 140 can provide a virtual machine or nonvirtual machine execution environment. By way of example, operating system 140 is provided by IBM's Virtual Machine/Enterprise Systems Architecture (VM/ESA) 1.1 operating system, except for the new features which comprise the present invention. The VM/ESA 1.1 operating system is described in a publication entitled VM/ESA General Information, which publication is available from International Business Machines Corporation at Mechanicsburg, Penna. by order number GC24-5550-02.
The operating system 140 includes a virtual memory manager subsystem 170 comprising a block page-out routine 150 (found in the VM/ESA 1.1 operating system) and a new slot allocation routine 155. One function of the virtual memory manager 170 is to move groups of related virtual pages from main storage 115 to DASD 160 in response to a request by another function in the operating system to free-up main storage. After receiving the request, the block page-out routine identifies a group of pages in virtual storage that currently reside in main storage and were accessed at approximately the same time. Then, the block page-out routine calls the slot allocation routine to obtain the optimum DASD slots to accomodate the group. Parameters in the call identify the total number of pages of the group and the address to store the DASD addresses allocated by the slot allocation routine. The criteria by which the slot allocation routine selects the optimum DASD slots are as follows: First, the number of discontiguous sets of contiguous storage slots on the DASD (a "slot" being just large enough to store one page of data) should be minimized because a different I/O operation and additional rotational movement are required for each discontiguous set. Preferrably, all of the selected DASD slots are contiguous so the block page-out routine can perform a single I/O operation to write out the entire group of pages. Otherwise, it will perform multiple I/O operations, one for each set of contiguous slots. Second, the one or more slots required to store this group of pages should be located relative to each other and to other subsequent groups such that the radial reader head movement and rotational movement required to access all the slots should be minimized. To this end, the slot allocation routine attempts to satisfy successive allocation requests from a localized region within a single extent. As noted above, an extent is a localized and contiguous storage area on the DASD which can be accessed with minimal radial movement of the reader head. Third, the allocation request should be satisfied, if possible, without leaving any small fragments or groups of slots which would not likely satisfy a subsequent request individually. To this end, during a first pass through an extent, the slot allocation routine searches for sets of contiguous available slots which either equal the allocation request leaving no fragment or which exceed the allocation request by so much that the remaining fragment would be large enough to satisfy most other subsequent requests. Also during the first pass, the slot allocation routine identifies a set, if any, of contiguous available slots which is larger than the allocation request and leaves a smaller fragment than any other set. If the first pass is unsuccessful in identifying an exact fit or a set which exceeds the allocation request by the large fragment, then the slot allocation routine selects the set, if any, of contiguous available slots which is larger than the allocation request and leaves the smallest fragment. If the first pass is unsuccessful in identifying any set which is larger than the allocation request, then the slot allocation routine makes a second pass of the same extent to identify multiple sets of contiguous available slots which together satisfy the allocation request. If such sets identified in the second pass cannot fully satisfy the allocation request, then the slot allocation routine continues with a pass through the next extent to complete the allocation.
The following is a more detailed description of the slot allocation routing and other aspects of the present invention. The DASD 160 is logically divided into multiple, active extents by an extent map 120 which is maintained in main storage and defines the beginning and ending real address of each active extent. An "active" extent is an extent that has one or more allocated slots. The extent map also includes a field for each active extent which specifies the percent to which the extent is currently utilized. As described in more detail below, the slot allocation routine references the extent map to determine which extent is a good candidate in which to search for available storage.
A bit map 125 is also maintained in main storage for each extent to map the availability of slots in the extent. Each bit map is partitioned such that a single bit in each part of the bit map represents a corresponding slot in the extent. The value of the bit indicates whether the slot is unavailable, i.e. currently allocated, or available. The extent map also describes those inactive extents (those which are currently completely available); however, to minimize storage, no bit map is provided for inactive extents. As described in more detail below, the slot allocation routine uses each bit map to generate a corresponding slot allocation buffer to expedite the search for available storage in the respective extent.
FIG. 2 illustrates the slot allocation buffer in detail. The slot allocation buffer comprises a header and set descriptors. Each set descriptor resides within the slot allocation buffer and describes a set of contiguous available slots in the extent. The header contains first and second pointers to the first and last set descriptors, respectively in the buffer which represent the first and last set of contiguous slots. If the first pointer is zero, then there are no available sets in the extent suitable for representation in the slot allocation buffer. It should be noted that the slot allocation routine classifies sets of different lengths into three categories. "Maxi-fragments" are sets whose length is so large that one large fragment can satisfy most allocation requests individually. "Midi-fragments" are sets with a shorter length but still long enough to many of the allocation requests individually . "Mini-fragments" are sets with still shorter length, and rarely can satisfy an allocation request individually. By way of example, one large fragment can store ten or more pages of data (with each page comprising 4096 bytes) and a large fragment having ten contiguous slots can satisfy seventy five percent of the allocation requests alone; one medium fragment can store four to nine pages of data; and and one small fragment can store one to three pages of data and a small fragment having three contiguous slots can satisfy five to ten percent of the allocation requests alone. Except when there are very few unused slots in any extent in the DASD, small fragments are not represented in the slot allocation buffer. This minimizes the search time within each slot allocation buffer, and omits nonoptimum sets. To further reduce the storage requirement and search time and to localize radial movement, each slot allocation buffer stores set descriptors only for a limited number of the sets within each extent. Other of the sets remain unrepresented in the slot allocation buffer until one or more of the set descriptors are fully utilized or utilized to the extent of leaving only a small fragment, thereby making room for set descriptors representing one or more other sets of contiguous available slots within the same extent. (Nevertheless, the bit map for each extent represents all of the slots in the extent.)
The first two bytes in each of the set descriptors indicate the real displacement from the start of the extent, the next byte indicates the length in bytes of the set, and the final byte points to the next set descriptor in the slot allocation buffer. Thus, the last byte in each set descriptor queues the set descriptors from one another (and the first pointer in the header identifies the first set descriptor in the queue). The slot allocation routine logically arranges the queue of set descriptors by ascending DASD address starting with the element designated by the first entry in the first row. Thus, the slots near to each other and the physical beginning of the extent are examined first in the selection process. As described in more detail below, this will improve the chances that sets of available contiguous slots are selected in order, either for the same allocation or successive allocations, and this further minimizes the requisite amount of seek time (i.e. radial head movement) for the same and subsequent allocations. The slot allocation routine uses the slot allocation buffer 175 to quickly identify the available sets of contiguous slots in the extent and the length of each set. With this information, the slot allocation routine can quickly identify the optimum contiguous sets of slots required to satisfy the request, thereby reducing the processing time required to move the pages to DASD. Also, each of the set descriptors is one fullword in length and all the set descriptors within one slot allocation buffer are stored in a continuous block, so the contents of an entire slot allocation buffer can be inspected with a single memory to cache transfer.
The following is a description of the allocation process. Initially, an operating system function notes that the free space in main storage has fallen below a predetermined level and alerts the block page-out routine. In response, the block page-out routine identifies a group of pages that are currently represented by virtual storage, actually stored in main storage and were accessed at approximately the same time. Then, the block page-out routine calls the slot allocation routine with a request to allocate the requisite number of slots in DASD to store the group of pages. The request also includes a main storage location into which the slot allocation routine should store the DASD slot addresses when allocated. With this information, the block page-out routine can subsequently direct the I/O subsystem 162 to physically write the data to DASD 160.
The allocation process of the present invention also minimizes overhead when a user/application program references one or more pages from the user/application program's virtual storage, and the referenced page(s) were previously written to DASD locators allocated by the slot allocation routine. In such a case, a page-in routine 158 within the virtual memory manager requests that the referenced page(s) and all other pages that were previously written out to DASD with the referenced page(s) be read into main storage. The reason for reading these other pages to main storage is the assumption that the user/application program will probably soon reference these other pages which are probably related to the referenced pages. Because the referenced page(s) and the other pages are likely stored in one set of contiguous slots, there is a minimum of I/O operators and radial read and rotational movement required during the reading process.
FIGS. 5-10 illustrate processing by the slot allocation routine. The slot allocation routine receives the allocation request (step 504 of FIG. 3), and sets a flag to indicate that a first pass search will commence through the current slot allocation buffer to identify optimum sets to satisfy the request (step 510). If this is the first allocation request since initial program load, then there is not yet a "current" slot allocation buffer (to satisfy the allocation request) or set descriptors within the slot allocation buffer (decision 514) or even a current extent. The selection of the current extent must precede the generation of the slot allocation buffer. To select the current extent, the slot allocation routine first considers the extent with the lowest relative address (decision 514 and step 910). The following describes the process of selecting a current extent when the first allocation request is made after initial program load, and subsequently, when a current extent cannot satisfy an allocation request. In step 910 (FIG. 7), the slot allocation routine first considers the extent with the lowest slot address as a candidate to satisfy the allocation request. To avoid unsuccessful searches, the slot allocation routine prefers to search extents which are substantially empty. Thus, only if the first extent is seventy percent or more empty is it immediatlely selected as the current extent for satisfying the allocation request (decision 914), then, a small fragment flat is reset to prevent representation of small fragments in the slot allocation buffer (step 940). Next, a branch is made to initialize the respective slot allocation buffer (step 950) as described below. Otherwise, the slot allocation routine determines if this extent is the emptiest extent yet encountered (which is always true for the 1st iteration) (decision 918) and its address is saved (step 922). If this extent is not the last extent on the DASD (decision 926), the next extent is considered as a candidate to satisfy the allocation request (step 930) and the iterative process beginning with step 914 is repeated. During the repeated iterations of this process, if any extents which are seventy percent or more empty are encountered, then such an extent is selected to satisfy the allocation request, and the iterations conclude with a branch to step 950. As described below, step 950 begins a subroutine for initializing the extent, i.e. generating a slot allocation buffer, as a prelude to the actual search which begins with decision 514. However, if no extents are encountered that are at least seventy percent empty and all the extents have been considered, the emptiest extent is selected to satisfy the allocation request (step 934). If this extent is at least twenty percent or more empty (decision 938), then a jump is made to the step 950. However, if this emptiest extent is less than twenty percent empty, then the small fragment flag is set so that small fragments can be selected (step 942), and a branch is made to the step 950.
In step 950, a cursor is initialized to the first bit in the bit map for this extent, and the header in the slot allocation buffer is set to zero to indicate there are currently no set descriptors queued. Then, the slot allocation routine calls a refill subroutine 954 illustrated in FIG. 9 providing the address is the slot allocation buffer for the first set descriptor. The refill subroutine is responsible for generating the slot allocation buffer set descriptors based on the corresponding bit map. In response to the call, the refill subroutine scans the corresponding bit map beginning with the bit designated by the bit map cursor and concludes this scan when a zero bit is found or the end of the bit map is reached (step 1010). The zero bit represents an available slot. If the scan reached the end of the bit map without finding a zero bit (decision 1014), the return code is set to nonzero to indicate failure and a return is made to the caller (step 1044). Otherwise, another scan is performed starting with the zero bit located by the previous scan (step 1018). It terminates when a non-zero bit is found or the end of a cylinder is reached. This second scan length represents the length of the set of contiguous available slots. This length is calculated by subtracting the zero bit location from the nonzero bit location (step 1026). If the flag indicating that the small fragments should be used is set (decision 1030) or the length is equal or greater than the medium fragment length (decision 1040), the beginning and ending locations and length of the set are stored as the current set descriptor (step 1034), the return code is set to zero to indicate success, and the refill subroutine returns to the caller. Otherwise (the length was too small), the process continues by branching to step 1010 to search the bit map for the next set. When the refill subroutine returns to the caller, if the return code is set to zero this means that the set descriptor was successfully filled. Then, the slot allocation routine queues this set descriptor as the last set descriptor in the slot allocation buffer (step 964). Next, the slot allocation routine determines if the slot allocation buffer is full (decision 970). If not, then another empty set descriptor is filled in another iteration of the refill subroutine 954 as above. This iterative process continues until either there are no more sets which meet the aforesaid criteria or the slot allocation buffer is completely filled. In either case, processing continues with step 514 in which the slot allocation buffer is searched to satisfy the allocation request. Referring again to decision 960, if the refill subroutine was not successful, a branch is made to step 514 to search those set descriptors that have been queued. Note that in this situation, the slot allocation buffer will not have the full complement of set descriptors. After selecting an extent to satisfy the allocation request and filling the corresponding slot allocation buffer, the slot allocation routine reads the first set descriptor (corresponding to a set), which set descriptor is identified by the first pointer in the header (step 522). During the first pass, only sets of contiguous available slots that are as large or larger than the allocation request will be considered thereby allowing one contiguous set to satisfy the entire allocation request so that the write operation can be performed with a single I/O operation (right and lower legs of decision 526). Consequently, if any of the set descriptors under consideration have a length which is less than the requested allocation, then the left leg of decision 526 leads to decision 810 which leads back to decision 542 which checks to see if this is the end of the queue. If not, the next set descriptor is selected (step 546).
Referring again to decision 526, if the length of the first set in the slot allocation buffer is exactly equal to the requested allocation, this set is immediately selected to satisfy the request (decision 526 and step 610 of FIG. 4). This is the ideal condition because one set can satisfy the request (requiring only one I/O operation) and there is no further fragmentation of available slots in the DASD. Pursuant to the selection, the bit map is updated to indicate that the slots within this set are no longer available (step 612), the selected row entry is dequeued (step 614) and the refill subroutine is called (step 618). The refill subroutine attempts to identify from the respective bit map a next large fragment or medium fragment set of contiguous available slots which is not yet represented in the slot allocation buffer for inclusion in the slot allocation buffer. If such a set exists (decision 622), it is queued at the end of the slot allocation buffer (step 626) because the set descriptors are logically arranged in order of physical address in the extent. Otherwise the slot allocation buffer remains partially empty, but still useful. Control is then returned to the block page-out routine with the real address of the allocation (step 558). In response, the block page-out routine orders the I/O subsystem to actually write the group of pages to DASD 160.
Referring again to decision 526 of FIG. 3, if the set length is greater than the requested allocation, the slot allocation routine determines if the set length exceeds the requested allocation by a large fragment, i.e. a length that is likely to satisfy most subsequent requests (decision 530). If the set length exceeds the allocation request by the large fragment, the first number of slots of the set which equal the allocation request are selected to satisfy the request (step 550). Then, the respective bit map is updated (step 552) to indicate that these slots are no longer available, and the beginning and length fields of the set descriptor are updated to represent the remaining large fragment (step 554). Then, control is returned to the block page-out routine with the real address of the allocation (step 558). In response, the block page-out routine orders the I/O subsystem 162 to actually write the group of pages into the allocated address on DASD 160.
Referring again to decision 530, if the current set length would not leave a large fragment, the slot allocation routine determines if this set leaves the smallest fragment. When this is the first set descriptor under consideration, in decision 534, the slot allocation routine determines that the current set would leave the smallest fragment, and therefore saves the address of this row entry for possible use later (step 538). Assuming there are additional set descriptors in this slot allocation buffer (decision 542), the slot allocation routine then examines the next set (step 546), and repeats the comparision to determine if this next set length equals the requested allocation (decision 526) or exceeds the requested allocation by a large fragment (decisions 526 and 530). If either decision is affirmative, then processing continues with step 610 or 550, respectively as described above because these two types of selections are preferrable to the other selection which will not leave a large fragment. Assuming neither decision is affirmative, the slot allocation routine determines if this next set length leaves a smaller fragment than the first set length (decision 534). If so, the address of this next set descriptor is saved at the expense of the first set descriptor (step 538). This loop is repeated until either decision 526 or decision 530 indicates a set whose length exactly equals the requested allocation or exceeds the requested allocation by a large fragment, or the last set descriptor is examined. When the last set descriptor in the slot allocation buffer has been examined and there have been no exact fits or set lengths that would leave a large fragment, the slot allocation routine determines if there is an entry for the set which would leave the smallest fragment (decision 710). If so, this set is made the current set descriptor (step 712). If the remaining fragment is a small fragment (decision 714), the slot allocation routine executes steps 610-630 in which the first blocks of the set which equal the requested allocation are selected to satisfy the request, the set descriptor is dequeued, and an attempt is made to replenish the slot allocation buffer. In this case, the small fragment will not now be represented by a set descriptor in the slot allocation buffer. This is an optimization to avoid the processing time that otherwise would be required to examine set descriptors for such small fragments when it is not likely that such a small fragment could ever satisfy an allocation request alone. It also avoids using such small fragments to satisfy requests that require multiple contiguous sets. However, if the remaining fragment is a medium fragment, then steps 550-554 are repeated to satisfy the request, update the respective bit map and update the beginning address and length information in the set descriptor of the slot allocation buffer.
Referring again to decision 710, if there is no entry for a set which exceeds the allocation request by a medium fragment or small fragment (and no set lengths that exactly equal the requested allocation or exceed the requested allocation by a large fragment) then there are no set lengths that can satisfy the allocation request alone. Consequently, the slot allocation routine begins the second pass through the same slot allocation buffer by setting a flag to indicate that this is no longer the first pass and two or more sets must be used to satisfy this request (step 715). Then, the slot allocation routine proceeds through step 522, decision 526, and decision 810 to step 814. In step 814, the slot allocation routine selects the first set descriptor in the queue to partially satisfy the requested allocation. Then the requested allocation is reduced by the length of the first set descriptor (step 818). Next, the slot allocation routine determines if the requested allocation has been satisfied and if there is a remainder which is greater than or equal to a medium fragment (decision 819). If not (which is the case for the first set descriptor which does not alone satisfy the request), the slot allocation routine updates the respective bit map to indicate that the slots represented by the first set descriptor are no longer available (step 820) and dequeues the first set descriptor (step 822). Next, the slot allocation routine calls the refill subroutine described above to search for and queue a next set from the same extent (step 826). If the refill is successful (decision 830), the slot allocation routine queues the new set at the end of the slot allocation buffer (step 834). Then, the slot allocation routine determines if the request has been satisfied (decision 840), and if not, continues with the next row entry (via step 522, decision 526 and decision 810) to try to satisfy the remaining allocation request. For each such set descriptor that is selected, the allocation request is reduced until it is finally satisfied. If a medium fragment or large fragment remains, the slot allocation routine branches to step 552 described above (decision 819). However, if the final set used to satisfy the allocation request exactly equals the remaining allocation request or leaves only a small fragment, then processing continues with step 558 as described above (decision 840).
Referring again to decision 830, if the refill subroutine is not successful and the request has not been satisfied (831), then the slot allocation routine continues with step 514 to utilize the next set in the current slot allocation buffer if one exists (decision 514, step 522 and decision 526) or select another extent beginning with step 910 to complete the allocation request if there are no more sets in the current extent. The routine beginning with step 910 is invoked whenever the current extent becomes empty of sets represented in the slot allocation buffer. The next extent which is described as a candidate for searching by the slot allocation routine to fulfill the present or a subsequent request is located adjacent to the previous extent to minimize radial movement of the read/write head.
Referring again to decision 831, if the request has been satisfied, the current set descriptor (which could not be successfully refilled) is left off the queue and processing continues with step 558.
Based on the foregoing, a computer system, process and programming embodying the present invention have been disclosed. However, numerous modifications and substitutions can be made without deviating from the scope of the present invention. For example, the slot allocation routine can allocate slots for records by a data base manager rather than a virtual storage manager as described above. Therefore, the present invention has been disclosed by way of illustration and not limitation, and reference should be made to the following claims to determine the scope of the present invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4528624 *||Mar 25, 1981||Jul 9, 1985||International Business Machines Corporation||Method and apparatus for allocating memory space based upon free space in diverse memory devices|
|US4536837 *||May 25, 1982||Aug 20, 1985||Elxsi||Improved disk file allocation and mapping system utilizing cylinder control blocks and file map having unbalanced tree structure|
|US4660130 *||Jul 24, 1984||Apr 21, 1987||Texas Instruments Incorporated||Method for managing virtual memory to separate active and stable memory blocks|
|US5021946 *||Aug 22, 1990||Jun 4, 1991||Modular Computer Systems, Inc.||Mostly contiguous file allocation technique involving file extension|
|US5200864 *||Jun 28, 1989||Apr 6, 1993||International Business Machines Corporation||Combining small records into a single record block for recording on a record media|
|US5247638 *||Jun 18, 1990||Sep 21, 1993||Storage Technology Corporation||Apparatus for compressing data in a dynamically mapped virtual data storage subsystem|
|US5247660 *||Jul 13, 1989||Sep 21, 1993||Filetek, Inc.||Method of virtual memory storage allocation with dynamic adjustment|
|AU5608890A *||Title not available|
|EP0061585A2 *||Feb 25, 1982||Oct 6, 1982||International Business Machines Corporation||Data processing apparatus with a storage device management facility|
|EP0406188A1 *||May 21, 1990||Jan 2, 1991||International Business Machines Corporation||Combining small records into a single record block for recording on a record media|
|1||Billiris, "An Efficient Database Storage Structure for Large Dyanmic Objects," Eighty International Conference on Data Engineering, 2-3 Feb. 1992 IEEE Computer Society Press pp. 301-308.|
|2||*||Billiris, An Efficient Database Storage Structure for Large Dyanmic Objects, Eighty International Conference on Data Engineering , 2 3 Feb. 1992 IEEE Computer Society Press pp. 301 308.|
|3||*||Reingold et al., Data Structures , Little, Brown & Company, (Boston, 1983) pp. 259 270.|
|4||Reingold et al., Data Structures, Little, Brown & Company, (Boston, 1983) pp. 259-270.|
|5||*||Widerhold, Database Design , McGraw Hill Book Company (New York, 1983) pp. 335 340.|
|6||Widerhold, Database Design, McGraw-Hill Book Company (New York, 1983) pp. 335-340.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US5555390 *||Jan 19, 1995||Sep 10, 1996||International Business Machines Corporation||Data storage method and subsystem including a device controller for respecifying an amended start address|
|US5574907 *||Nov 30, 1994||Nov 12, 1996||Microsoft Corporation||Two-pass defragmentation of compressed hard disk data with a single data rewrite|
|US5623701 *||Jun 6, 1995||Apr 22, 1997||International Business Machines Corporation||Data compression method and structure for a direct access storage device|
|US5666560 *||Aug 3, 1995||Sep 9, 1997||International Business Machines Corporation||Storage method and hierarchical padding structure for direct access storage device (DASD) data compression|
|US5682499 *||Jun 6, 1995||Oct 28, 1997||International Business Machines Corporation||Directory rebuild method and apparatus for maintaining and rebuilding directory information for compressed data on direct access storage device (DASD)|
|US5732256 *||Aug 30, 1995||Mar 24, 1998||Microsoft Corporation||CD-ROM optimization and stream splitting|
|US5745749 *||Jun 21, 1995||Apr 28, 1998||International Business Machines Corp.||Method and system of file version clustering of object blocks using a compiler and database and having a predetermined value|
|US5848405 *||Apr 21, 1997||Dec 8, 1998||Oracle Corporation||Method and apparatus for identifying new data by address ranges|
|US5860132 *||Dec 19, 1996||Jan 12, 1999||International Business Machines Corporation||Storage dispersal system based on related, shared data storage space reservation|
|US6065095 *||Sep 16, 1999||May 16, 2000||Western Digital Corporation||Method for memory allocation in a disk drive employing a chunk array and identifying a largest available element for write caching|
|US6094707 *||Sep 16, 1999||Jul 25, 2000||Western Digital Corporation||Method for memory allocation in a disk drive employing a chunk array and identifying a smallest available element for write caching|
|US6175900 *||Feb 9, 1998||Jan 16, 2001||Microsoft Corporation||Hierarchical bitmap-based memory manager|
|US6215748 *||Apr 9, 1997||Apr 10, 2001||Sony Corporation||Data storage with data masking array, apparatus and method therefor|
|US6578120 *||Jun 24, 1997||Jun 10, 2003||International Business Machines Corporation||Synchronization and resynchronization of loosely-coupled copy operations between a primary and a remote secondary DASD volume under concurrent updating|
|US6591334 *||Jun 19, 2000||Jul 8, 2003||Hitachi Global Storage Technologies Netherlands B.V.||Method and apparatus for reducing space allocation failures in storage management systems|
|US6640290||Nov 28, 2000||Oct 28, 2003||Microsoft Corporation||Easily coalesced, sub-allocating, hierarchical, multi-bit bitmap-based memory manager|
|US6842834 *||Mar 20, 2003||Jan 11, 2005||International Business Machines Corporation||Synchronization and resynchronization of loosely-coupled copy operations between a primary and a remote secondary DASD volume under concurrent updating|
|US7143262 *||Jun 27, 2003||Nov 28, 2006||Hitachi, Ltd.||Method, apparatus and program for allocating storage area to virtual volume|
|US7191282 *||Jul 21, 2003||Mar 13, 2007||Autodesk Canada Co.||Data processing apparatus and method for determining the amount of free storage space in such an apparatus|
|US7660837 *||Oct 6, 2006||Feb 9, 2010||Broadcom Corporation||Method for automatically managing disk fragmentation|
|US7765244||Oct 6, 2006||Jul 27, 2010||Broadcom Corporation||Fast and efficient method for deleting very large files from a filesystem|
|US7797485 *||May 8, 2007||Sep 14, 2010||Samsung Electronics Co., Ltd.||Method and apparatus for allocating disc space for recording files|
|US7860896||Jan 12, 2010||Dec 28, 2010||Broadcom Corporation||Method for automatically managing disk fragmentation|
|US7966351||Jul 26, 2010||Jun 21, 2011||Broadcom Corporation||Fast and efficient method for deleting very large files from a filesystem|
|US8607009 *||Jul 13, 2006||Dec 10, 2013||Microsoft Corporation||Concurrent virtual machine snapshots and restore|
|US8984244 *||Nov 13, 2013||Mar 17, 2015||Microsoft Technology Licensing, Llc||Concurrent virtual machine snapshots and restore|
|US9619157 *||Apr 3, 2014||Apr 11, 2017||Analysis Solution Llc||High-speed data storage|
|US20030177322 *||Mar 20, 2003||Sep 18, 2003||Crockett Robert Nelson||Synchronization and resynchronization of loosely-coupled copy operations between a primary and a remote secondary DASD volume under concurrent updating|
|US20040057200 *||Jul 21, 2003||Mar 25, 2004||Autodesk Canada Inc.||Data processing apparatus|
|US20040098537 *||Jun 27, 2003||May 20, 2004||Kazuyoshi Serizawa||Method, apparatus and program for allocating storage area to virtual volume|
|US20070271436 *||May 8, 2007||Nov 22, 2007||Samsung Electronics Co., Ltd.||Method and apparatus for allocating disc space for recording files|
|US20080005205 *||Oct 6, 2006||Jan 3, 2008||Broadcom Corporation||Fast and efficient method for deleting very large files from a filesystem|
|US20080005206 *||Oct 6, 2006||Jan 3, 2008||Broadcom Corporation||Method for automatically managing disk fragmentation|
|US20080022032 *||Jul 13, 2006||Jan 24, 2008||Microsoft Corporation||Concurrent virtual machine snapshots and restore|
|US20080229045 *||Mar 16, 2007||Sep 18, 2008||Lsi Logic Corporation||Storage system provisioning architecture|
|US20100290755 *||Jul 26, 2010||Nov 18, 2010||Broadcom Corporation||Fast and efficient method for deleting very large files from a filesystem|
|US20110153929 *||Dec 22, 2009||Jun 23, 2011||International Business Machines Corporation||Disk memory utilization management using available slot clusters|
|US20140075141 *||Nov 13, 2013||Mar 13, 2014||Microsoft Corporation||Concurrent virtual machine snapshots and restore|
|US20150220452 *||Feb 27, 2014||Aug 6, 2015||Lsi Corporation||System, Method and Computer-Readable Medium for Dynamically Mapping a Non-Volatile Memory Store|
|US20150286420 *||Apr 3, 2014||Oct 8, 2015||ANALYSIS SOLUTION LLC, dba Gearbit||High-Speed Data Storage|
|U.S. Classification||711/112, 711/172, 707/999.205, 707/999.202|
|International Classification||G06F12/00, G06F3/06, G06F12/02|
|Cooperative Classification||Y10S707/99953, Y10S707/99956, G06F3/0601, G06F2003/0697|
|Jun 15, 1992||AS||Assignment|
Owner name: INTERNATIONAL BUSINESS MACHINES CORPORATION, A COR
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST.;ASSIGNORS:BLANDY, GEOFFREY O.;ROSSI, CATHERINE M.;REEL/FRAME:006161/0810
Effective date: 19920615
|Jun 19, 1998||FPAY||Fee payment|
Year of fee payment: 4
|Sep 3, 2002||REMI||Maintenance fee reminder mailed|
|Feb 14, 2003||LAPS||Lapse for failure to pay maintenance fees|
|Apr 15, 2003||FP||Expired due to failure to pay maintenance fee|
Effective date: 20030214